Gene expression profiling with microarrays and in vivo deuterated water labeling in patients with chronic lymphocytic leukemia (CLL) have shown that tumor activation and proliferation predominantly occur in the lymph node (LN). It is thought that CLL cells actively shape the surrounding microenvironment to support tumor survival and growth.
To further dissect the biology of CLL and its interactions with the tumor microenvironment, we performed total RNA sequencing of purified tumor cells from paired peripheral blood (PB) and LN in 29 patients. Samples separated by anatomic compartment in principal component analysis. A total of 883 protein-coding genes were differentially expressed (DE) between PB and LN (fold-change >2, Q <0.1). Among these DE genes, 82% were upregulated and 18% were downregulated in LN compared to PB. Gene Set Enrichment Analysis using experimentally-derived lymphoid gene signatures (https://lymphochip.nih.gov/signaturedb) revealed activation of B-cell receptor (BCR), NF-κB, JAK-STAT, cytokine and NOTCH signaling as well as increased proliferation and metabolism in LN (normalized enrichment score >1.6, Q <0.05). In contrast, PB tumor cells were characterized by a quiescent transcriptional program.
Long non-coding RNAs (lncRNAs) are a novel class of transcripts with emerging significance in cancer. We detected 134 DE lncRNAs between PB and LN. To gain insight into the function of lncRNAs, DE lncRNAs and protein-coding genes were subjected to weighted gene co-expression network analysis grouping highly co-expressed transcripts into modules. One of 5 identified modules contained protein-coding genes representing BCR activation and cell proliferation. Among 36 lncRNAs in this module, CCAT1 was overexpressed in LN-resident CLL cells. Overexpression of CCAT1 has also been reported in acute myeloid leukemia and some solid tumors, in which CCAT1 regulates MYC transcription. While the function of most lncRNAs remains unknown, our data suggest a network of lncRNAs may regulate survival pathways and proliferation of CLL cells in the microenvironment.
Next, we profiled the transcriptome of bulk LN samples from 33 patients with CLL and 4 normal donors to characterize the cellular composition of CLL-LNs. We applied CIBERSORT (Newman, Nat Methods 2015) to infer the relative proportion of leukocyte subsets. First, B cells were more abundant in CLL than normal LNs (median 66% versus 34% of cellular elements, Q < 0.0001). Among non-B cells, CLL-LNs had more γδ-T cells (median 12% versus 1%, Q = 0.04), but fewer CD4+ T cells (median 34% versus 57%, Q = 0.008), NK cells (median 0% versus 4%, Q = 0.008), and mast cells (median 0% versus 1%, Q = 0.01) than normal LNs. No difference was observed in the frequency of CD8+ T-cells or monocyte-macrophage lineage cells between CLL and normal LNs. Because M2-polarized tumor-associated macrophages have been implicated in tumor development, we also assessed the ratio of M1 to M2 macrophages. However, macrophages from CLL-LNs did not exhibit increased M2 polarization as compared to normal LNs.
In summary, we confirmed BCR and NF-κB activation reported in prior microarray studies and additionally identified the upregulation of JAK-STAT and NOTCH signaling in LN-resident tumor cells. In addition, network analysis suggested a possible role for lncRNAs in the response of CLL cells to microenvironmental stimuli. Significant alterations in the cellular composition of CLL-LNs included the relative expansion of B cells and γδ-T cells. Our data highlight the complexity of tumor-microenvironment interactions in CLL and reveal novel aspects that merit further investigation.
This research was supported by the Intramural Research Program of the NHLBI.
Wiestner: Acerta Pharma: Research Funding; Pharmacyclics: Research Funding.
Asterisk with author names denotes non-ASH members.